About one year ago, Datadog published a report and statistics about serverless usage. This report has been updated with interesting figures, for example the median lambda invocation duration decreased by half in a year. Or surprisingly, 57% of the cloudformation stacks include only one lambda function... Why would you do that?!

I love this kind of feedback from customers, it gives so much useful insights: event ordering, idempotency, CI/CD, ... Also, this recent AWS blog post ("Mistakes to Avoid When Implementing Serverless Architecture with AWS Lambda") complements it perfectly: monolithic functions, synchronous architectures, shared code, ... 2 must read!

Have you ever dreamt of using the same docker image both in ECS and Lambda? This article explains how, and also how they've used Lambda in addition to ECS when traffic was bursting.

Now that most of us can get a vaccine against covid-19, it may be interesting to automatically check if there are some availabilities, instead of doing it manually every day. This article explains how to implement such a solution, using Lambda and Cloudwatch Events.

Interesting article from Yan Cui on the way to use Lambda layers: certainly not to replace npm, maven or pip! Dependency managers are indeed much simpler than layers. But then what? Use layers for deployment. Yan tells you...

When building and deploying serverless application today, you have the choice between multiple frameworks. If you wonder which one to use for your next application, this article compares Serverless Framework, AWS SAM and AWS CDK, for the exact same architecture. No real winner, as always it depends :)

Many people and companies fear lock-in when talking about serverless. But what if there was a solution to avoid that? Hexagoxal Architecture may be this one. In this great article, Filippo Cilluffo, details how to do it with a Go sample. Simply brilliant!

Not an easy topic: do you know in which cases your function retries, and when you can use Dead Letter Queuest? This article is a great summary of the different cases. To keep somewhere as a reference...

Once again, Yan Cui delivers a superb article. Very detailed, it goes through cold starts and how to reduce them, timeouts and how to monitor and mitigate them, and how to debug slow Lambda response times.
And if you're interested in optimizing performance, have a look at this 3-part (part 1 / part 2 / part 3) blog post from James Beswick on AWS.
If that's not enough, you can also read that one: "4 Tips for AWS Lambda Optimization in Production". With that, you don't have any excuse to provide poor-performance Lambda functions!

If you are reading this newsletter, there are chances that you already know serverless quite a bit. But if you start or want to share a great introduction with a colleague, this series of articles (part 1, part 2, part 3) from Benjamin Smith is pretty good.

Interesting story of Dylan O'Reagan, although a little wordy, on their path to serverless: What and more important, why did they choose one technology over another. Java is just one choice among others: Infra As Code, Queue or not queue? API Gateway or Load Balancer, ...

If you also choose to go serverless with Java, this article might be of interest. Daniel Oh describes how to create Lambda functions using Quarkus, which is a bit faster than Spring. And that's a good idea to avoid cold starts...

This article describes a very common thing you will have to do if you provide an authenticated API (with API Gateway and Lambda): protect it with Cognito. Very good article from Artem Lovan.

AWS updates

Few updates to notice on the serverless landscape:

• Step Functions Data Flow Simulator. AWS Step Functions may sometimes be complicated, especially when dealing with input, output of each state. This new feature in the console will probably help.
• Grafana to monitor your Lambda Functions. Grafana is quite standard today to monitor systems. Now, you can also use it to monitor your Functions.
• CloudFront Functions. You probably know Lambda@Edge, which permits you to run your functions at the edge, to manipulate HTTP requests and responses. Now it becomes even faster with Cloudfront Functions.
• Integration of EventBridge in StepFunctions. This is probably my favorite one. You can now send events directly from your state machine. Really cool to implement choreography and orchrestration patterns for your microservices.
• AWS Lambda Extensions Are Now Generally Available.
• AWS App Runner. This is actually my favorite one, even if it is not serverless. App Runner permits you to create a web application or backend APIs in a matter of minutes. Looks really cool.s